Multimodal Large Language Models (MLLMs) have shown remarkable progress in temporal or spatial localization tasks, but struggle with joint spatio-temporal video grounding (STVG). We identify two key bottlenecks hindering this capability: (1) the sheer number of visual tokens makes long-range and fine-grained visual modeling challenging; (2) generating a long sequence of bounding boxes in text makes it hard to accurately align each box with its specific video frame. Distinct from prior efforts that rely on attaching complex modules, we argue for a more elegant paradigm that unlocks the inherent potential of MLLMs and leverages their strengths. To this end, we propose \textbf{\textit{SpaceVLLM}}, a MLLM equipped with spatio-temporal video grounding capabilities. Specifically, we propose Spatio-Temporal Aware Queries, interleaved with video frames, to guide the MLLM in capturing both static appearance and dynamic motion features. We further present a lightweight Query-Guided Space Head that maps queries to precise spatial coordinates, bypassing the need for direct textual coordinate generation and enabling the MLLM to focus on video understanding. To further facilitate research in this area, we propose an automated data synthesis pipeline to construct \textbf{V-STG} dataset, comprising 110K STVG instances. Extensive experiments show that \textit{SpaceVLLM} achieves the state-of-the-art performance on STVG benchmarks and maintains strong performance on various video understanding tasks, validating our approach's effectiveness.

Large-scale, high-quality data is essential for advancing the reasoning capabilities of large language models (LLMs). As publicly available Internet data becomes increasingly scarce, synthetic data has emerged as a crucial research direction. However, existing data synthesis methods often suffer from limited scalability, insufficient sample diversity, and a tendency to overfit to seed data, which constrains their practical utility. In this paper, we present \textit{\textbf{GRIP}}, a \textbf{G}raph-based \textbf{R}easoning \textbf{I}nstruction \textbf{P}roducer that efficiently synthesizes high-quality and diverse reasoning instructions. \textit{GRIP} constructs a knowledge graph by extracting high-level concepts from seed data, and uniquely leverages both explicit and implicit relationships within the graph to drive large-scale and diverse instruction data synthesis, while employing open-source multi-model supervision to ensure data quality. We apply \textit{GRIP} to the critical and challenging domain of mathematical reasoning. Starting from a seed set of 7.5K math reasoning samples, we construct \textbf{GRIP-MATH}, a dataset containing 2.1 million synthesized question-answer pairs. Compared to similar synthetic data methods, \textit{GRIP} achieves greater scalability and diversity while also significantly reducing costs. On mathematical reasoning benchmarks, models trained with GRIP-MATH demonstrate substantial improvements over their base models and significantly outperform previous data synthesis methods.

Phenotypic variation results from the combination of genotype, the environment, and their interaction. The ability to quantify the relative contributions of genetic and environmental factors to complex traits can help in breeding crops with superior adaptability for growth in varied environments. Here, we developed and extensively evaluated the performance of an explainable machine-learning framework named explainable genotype-by-environment interactions prediction (EXGEP) to accurately predict the grain yield in crops. To assess the performance of EXGEP, we applied it to a dataset comprising 70 693 phenotypic records of grain yield traits for 3793 hybrids (also including both genotype and environmental condition data). When used with four different combinations of genotypes and environmental data, EXGEP exceeded the yield prediction performance of the classic model Bayesian ridge regression model by 17.37%-42.35%. Moreover, EXGEP incorporates SHapley Additive exPlanations values that can uncover complex nonlinear relationships between genotype and environment and identify key features, and their interactions, that provide the main contributions to model performance, thus enhancing our understanding of genotype-by-environment interactions. Additionally, data from a series of tests support that EXGEP exhibits superior performance in terms of prediction accuracy and explainability. Our development of EXGEP and comparisons of it against alternative models provides valuable insights into methods for accurately predicting complex traits in multiple environments.

This paper considers high reliability transmission for anchored indoor New Radio Unlicensed (NR-U) systems, in which unlicensed spectrum with wideband operation is combined with other NR licensed spectrum as anchor. To achieve high reliable transmission using unlicensed spectrum, success transmission ensuring scheme is investigated for indoor NR-U deployment. By modeling of Success Transmission Probability (STP) with nonlinear Levenberg-Marquardt method, STP-based packet request algorithm and STP-based channel allocation algorithm are proposed for initial transmission and Automatic Repeat reQuest (ARQ) retransmission, respectively. Simulation results show that the proposed success transmission ensuring scheme significantly decreases retransmission ratio and further provides less transmission latency and higher throughput. In a word, this paper provides an effective solution which satisfies high reliable transmission requirement for NR-U.

Eukaryotic genomes are organized in a three-dimensional spatial structure. In this regard, the development of chromosome conformation capture methods has enabled studies of chromosome organization on a genomic scale. Hi-C, the high-throughput chromosome conformation capture method, can reveal a population-averaged, hierarchical chromatin structure. The typical Hi-C analysis uses a two-dimensional (2D) contact matrix that indicates contact frequencies between all possible genomic position pairs. Oftentimes, however, such a 2D matrix is not amenable to handling quantitative comparisons, visualizations and integrations across multiple datasets. Although several one-dimensional (1D) metrics have been proposed to depict structural information in Hi-C data, their effectiveness is still underappreciated. Here, we first review the currently available 1D metrics for individual Hi-C samples or two-sample comparisons and then discuss their validity and suitable analysis scenarios. We also propose several new 1D metrics to identify additional unique features of chromosome structures. We highlight that the 1D metrics are reproducible and robust for comparing and visualizing multiple Hi-C samples. Moreover, we show that 1D metrics can be easily combined with epigenome tracks to annotate chromatin states in greater details. We develop a new framework, called HiC1Dmetrics, to summarize all 1D metrics discussed in this study. HiC1Dmetrics is open-source (github.com/wangjk321/HiC1Dmetrics) and can be accessed from both command-line and web-based interfaces. Our tool constitutes a useful resource for the community of chromosome-organization researchers.

Plug-in Electric Vehicles (PEV) are characterized as a type of unconventional electric load, and their Fast Charging Stations (FCS) are capital-intensive transportation service infrastructure. Therefore, FCS location planning must consider the requirements of electricity and transportation infrastructures simultaneously. This paper, for the first time, proposes a graph-computing based integrated location planning model, which maximizes PEV charging convenience while ensuring the power grid's reliability. In addition, the proposed model captures the long-term costs of critical grid assets induced by uncertainty and impulsiveness of charging demand. Finally, the model can be easily scaled to various coupling configurations and temporal resolutions through a graph-based scheme. The proposed model is cast as a multi-objective mixed-integer problem and is solved by the cross-entropy (CE) optimization algorithm, in which the computational efficiency is significantly improved with graph parallel computing techniques. The proposed planning model and graph implementation method are validated on a synthetic power-transportation coupled network.

In this paper, the resource allocation for vehicle-to-everything (V2X) underlaying 5G cellular mobile communication networks is considered. The optimization problem is modeled as a mixed binary integer nonlinear programming (MBINP), which minimizes the interference to 5G cellular users (CUs) subject to the quality of service (QoS), the total available power, the interference threshold, and the minimal transmission rate. To achieve that, the original MBINP is decomposed into three steps: transmission power initialization, subchannel assignment, and power allocation. Firstly, the minimum transmission power required by the V2X users (VUs) is set as the initial power value. Secondly, the Hungarian algorithm is used to obtain the appropriate subchannel. Finally, an optimization mechanism is proposed to the power allocation. Simulation results show that the proposed algorithm can not only ensure the minimal transmission rate of VUs but also further improve the CUs’ channel capacity under the premise of guaranteeing the QoS of the CUs.

In this paper, we present a method to combine a grammatical model that encodes a priori shape information with the ziplock snakes presented by Neuenschwander et al. A competing mechanism is adopted to take advantage of the shape models without inducing excessive computation. The resulting model-based ziplock snakes have many advantages over the original model: they can accurately locate contour features, produce more refined results, and deal with multiple contours, missing image cues, and noise.

Active contour models, or snakes, are effective and robust in contour extraction. In most papers on snakes, an initialization close to the desired contour is assumed to be provided, which is inappropriate in many cases. The ziplock snake model presented by Neuenschwander et al. (1997), however, needs only two user-supplied endpoints. The optimization process for a ziplock snake starts from the two endpoints and progresses towards the center of the snake. In this paper, we present a method to combine a grammatical model that encodes a priori shape information with the ziplock snakes. A competing mechanism is adopted to take advantage of the shape models without inducing excessive computation. The resulting model-based ziplock snakes have many advantages over the original model: They can accurately locate contour features, produce more refined results, and deal with multiple contours, missing image cues and noise.

The original snake models require a close initialization which in many situations are difficult to acquire. The balloon model presented by Cohen et al. to solve this problem suffers from the difficulty of choosing a constant inflating force due to variable internal shrinking forces and non-constant boundary intensity levels. Xu et al., on the other hand, proposed to use a pressure force to exactly offset the shrinking forces. The resulting model achieves better stability in terms of parameter insensitivity by sacrificing smoothness constraints, thus it would go through even small gaps on a boundary. We instead propose to compute an adaptive inflating force locally for each snaxel so that it is just enough to overcome the image force. A new smoothness constraint which can maintain smoothness without any shrinking side-effects is also presented, along with a new way to resample a balloon without significantly reducing its tension. The combined model is sensitive to weak and incomplete boundaries, and yet able to overcome noise edges. Experimental results are reported to support our statements.

Segmentation of ultrasound images is difficult due to the existence of speckle noise. Erroneous edges from speckle noise are not only abundant but also have large magnitude due to the multiplicative nature of speckle noise. Moreover, boundary edges are usually incomplete, being missing or weak at some places. We propose a system to address these problems in two steps. First, based on the observation that boundaries in ultrasound images have the appearance of straight or gently curving line segments, we adopt Sha'ahsua and Ullman's (1988) saliency map method to reduce speckle noise and enhance edges. Then we use a new snake model, which we call a systolic snake, to perform a multi-level feature search. The systolic snake can not only overcome local minima, but also effectively use both strong and weak image information. Furthermore, the system can be used in an automatic system since, unlike other snake models, ours does not need a close initialization The resulting system is tested on some ultrasound loin images and results are promising.